Blue Waters: A Super System for Highly Productive Frontier Science -- Experiences and Lessons Learned
Dr. William T.C. Kramer | National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign
The Blue Waters system is the first general purpose, open science, sustained-petaflop supercomputer. It is a powerful resource for the nation’s open-science researchers. Blue Waters is a ten-year project with an overall cost of over $500M and went into full service in April 2013 and supports thousands of researchers doing Frontier Science and Engineering computation and data analysis that is not possible any other way.
This talk begins with a brief overview of the Blue Waters project, a Leadership-class system, and how Blue Waters is used to support extreme-scale Frontier Science. I will discuss how the system and the science being done on it has evolved over the past four years. I will then discuss lessons learned from Blue Waters and highlight a very few of the best practices that have helped the Blue Waters Project excel. We will briefly explain why Blue Waters publically refused to be listed on the Top 500 list. The talk will conclude by looking to the future regarding how the experience gained with Blue Waters will help guide the delivery of future sustained performance Leadership Class technologies.
BIO: Under Kramer’s guidance, the Blue Waters publically declined to be listed on the Top 500 list to make the point that the list does not correlate to the actual sustained performance of large-scale systems. In addition to Blue Waters Director, Kramer is a Research Professor of Computer Science in the Computer Science department at UIUC, and is currently the PI of the NSF funded Global Initiative to Enhance @scale and distributed Computing and Analysis Technologies (GECAT) project, and the DOE/ASCR funded Holistic Measurement Driven Resiliency HMDR award that studies failure and recovery for exascale systems.